OneResearchClaw

OneResearchClaw takes real-world material in many formats, meeting audio or video, documents, tables, slide decks, ZIP archives of mixed material, or links to arXiv papers, YouTube or Bilibili videos, and runs it through a pipeline that grounds the content, researches the relevant literature, reviews the output for quality, and exports the result in multiple formats.

The pipeline can identify and split multiple topics within one input, such as a single meeting covering several subjects, and produce independent reports for each. Research depth is adjustable between simple, medium and complex settings to trade off speed against thoroughness, and a review-rewrite loop applies bounded iterative revision to improve output quality. A skill-evolution feature lets the system save personalized preferences from user feedback as selectable derived versions of its skills over time.